Skip to main content
Transformidy

Article

Customers Now Expect AI to Show Its Work

Demand for AI transparency in customer interactions rose 63% year over year. 95% of consumers now expect to understand the 'why' behind an automated decision. The reasoning-model era normalized showing work. Customer expectations followed it home.

Published
September 20, 2024
Updated
June 18, 2026
Reading time
7 min
Editorial illustration for Customers Now Expect AI to Show Its Work

2026 updated analysis

What changed since the original article

This page keeps the original Transformidy article as the canonical record and leads with the current interpretation, source notes, and Revenue Unknown framing.

Where the Expectation Came From

Zendesk's 2026 CX Trends research found that 95% of consumers now expect to understand the "why" behind an automated decision, with demand for greater AI transparency rising 63% year over year. That is not a slow-building sentiment; it is a sharp, recent shift in what customers consider a baseline requirement for interacting with automated systems.

The timing lines up with a specific technology shift. Reasoning-capable AI models, the kind that visibly work through a problem step by step before producing an answer, moved from a research novelty into mainstream, everyday tools over the past two years. Once customers began seeing AI systems show their reasoning in one context, an unexplained decision in another context, a denied insurance claim, a flagged transaction, a rejected application, started to read less like an inherent limitation of the technology and more like something being deliberately withheld.

That reframing is the mechanism worth understanding. A customer does not need to know the technical difference between a reasoning model and a black-box classifier to feel the difference between an explained decision and an unexplained one. Once the explained version becomes a normal experience anywhere, its absence elsewhere starts to feel like a choice, not a constraint.

Zendesk 2026 CX Trends research
63%

Year-over-year rise in demand for AI transparency, alongside 95% expecting an explanation for automated decisions

A fast-moving expectation shift, not a gradual accumulation of dissatisfaction.

Accurate and Trusted Are Different Tests

The line Zendesk's research uses to frame this finding is worth sitting with directly: "It's clarity that drives trust and loyalty in CX engagements." That sentence deliberately does not say accuracy. A business can build an AI system that makes the objectively correct decision in the overwhelming majority of cases, and still fail the trust test if the customer on the receiving end of any single decision cannot see why it was made.

This is a genuinely different design requirement than the ones most AI customer service investment has focused on. Resolution rate, accuracy, and speed all measure whether the system got the right answer and delivered it efficiently. None of them measure whether the customer understood how the system arrived there. A system can score well on every one of those metrics and still generate the exact trust erosion Zendesk's 63% figure describes, if it never surfaces its reasoning to the person affected by it.

The organizations best positioned here are not necessarily the ones with the most accurate AI, but the ones that treated explainability as a first-class design requirement from the start, not a compliance afterthought bolted on after a complaint. Given how quickly this expectation is rising, closing the gap later, after a wave of trust-erosion incidents rather than ahead of them, is a materially worse position to be building from.

What changes

Optimizing for Accuracy vs. Optimizing for Clarity

One measures whether the system was right. The other measures whether the customer could tell.

Accuracy-only systems: correct decisions delivered without visible reasoning, meeting a technical bar customers no longer treat as sufficient on its own. Clarity-first systems: the same decisions delivered with a specific, understandable reason attached, meeting the trust bar 95% of consumers now bring to any automated interaction.

"It's clarity that drives trust and loyalty in CX engagements."

Zendesk, 2026 CX Trends research

The Leadership Move

The structural choice for any organization deploying customer-facing AI is whether to treat explainability as a design requirement from the outset, or as a feature to retrofit once the trust cost of opaque decisions becomes visible in churn or complaint data.

Ownership

Product and CX leadership jointly own the decision to require reasoning-visible design in customer-facing AI systems, rather than leaving explainability as an engineering nice-to-have subordinate to accuracy and speed metrics alone.

Tradeoff

Building AI systems that can surface their reasoning to customers takes additional design and engineering investment beyond optimizing for raw accuracy or resolution speed. The tradeoff against skipping that investment is a widening gap, at a 63% year-over-year growth rate, between what customers expect and what the system delivers.

Human consequence

A customer denied a claim, flagged on a transaction, or rejected on an application by an opaque AI system experiences that outcome as arbitrary, regardless of whether the underlying decision was correct, while the same outcome delivered with a clear reason is experienced as fair, even when unwelcome.

Next Move

If your organization deploys AI for high-stakes customer decisions: Audit whether customers currently receive any specific reasoning for denials, flags, or rejections, and treat closing that gap as a trust investment, not just a technical one.

If you are evaluating a new AI vendor or system: Add explainability as an explicit selection criterion alongside accuracy and speed, since Zendesk's data suggests customers increasingly weight it as heavily as the outcome itself.

FAQ

How many consumers expect an explanation for an AI-made decision?

According to Zendesk's 2026 CX Trends research, 95% of consumers now expect to understand the "why" behind an automated decision. Demand for AI transparency specifically rose 63% year over year, indicating this expectation is accelerating, not holding steady.

Why did this expectation rise so quickly?

Reasoning-capable AI models becoming mainstream in customer-facing tools normalized the idea that an AI system can show its work, not just produce an answer. Once customers experienced tools that explained their reasoning, an unexplained automated decision, a denied claim, a rejected order, a flagged transaction, started to feel like a withheld answer rather than a normal limitation of the technology.

Does accuracy alone satisfy this expectation?

No. Zendesk's research frames the finding specifically around clarity, not correctness: the quoted conclusion is that "it's clarity that drives trust and loyalty in CX engagements," distinct from whether the underlying decision was accurate. A correct decision delivered without explanation can still damage trust if the customer cannot see why it was made.

What should a business do if its AI systems cannot currently explain their reasoning?

Treat explainability as a customer experience requirement, not just a technical or compliance one, and prioritize it in AI vendor selection and internal system design. A 63% year-over-year rise in transparency demand suggests this gap will keep widening for any organization that does not close it deliberately.

Sources & References

Original article archive

Original article published September 20, 2024: "OpenAI o1 Model and Its Impact on CX ". Preserved here for provenance, historical context, and citation continuity.

The release of OpenAI's GPT-4 and o1 Model, has brought forth transformative changes for businesses. Using these technologies, organizations can evolve their chatbot capabilities, offering new tools, and businesses can leverage these advancements to refine customer experience strategies, boost engagement, drive revenue, and enhance partnerships. This Transformidy insight dives deeper into the changes, addresses hallucination concerns, and expands into their customer experience use cases.

OpenAI o1 Models: Key Features

The OpenAI o1 models bring new levels of efficiency, flexibility, and interactivity to AI applications, offering businesses a more refined toolkit for customer-facing and internal processes. Here are some key features:

  1. Greater Customization and Control: OpenAI o1 models allow for even more extensive customization, enabling businesses to fine-tune the AI’s behavior based on brand needs. These models can be adapted to follow specific instructions more closely, ensuring that every interaction remains on-brand and personalized to the company's unique customer service style.
  2. Faster Response Times: Leveraging cutting-edge architecture, o1 models offer significantly faster processing times, reducing latency during customer interactions. This quickness improves the overall user experience and provides instantaneous responses to customer queries, ensuring seamless real-time interactions across channels.
  3. Enhanced Conversational Context: Building upon GPT-4’s improvements in context retention, OpenAI o1 models can manage extended and complex conversations with greater accuracy, making them ideal for customer service scenarios where dialogue continuity is crucial. This advancement empowers businesses to deliver more nuanced and meaningful interactions, fostering stronger customer relationships.
  4. Superior Multi-modal Capabilities: Like GPT-4, o1 models continue to push the boundaries with multi-modal interactions, allowing companies to create more dynamic customer experiences. Customers can engage through text, voice, or images, which opens new avenues for businesses to deliver more comprehensive services.
  5. Advanced-Data Integration: The OpenAI o1 models provide enhanced integration with third-party systems, such as CRM platforms, loyalty programs, and analytics dashboards. This makes it easier for businesses to extract actionable insights from customer data, enabling real-time personalization and better-informed decision-making.
https://www.youtube.com/watch?v=3k89FMJhZ00
OpenAI and its o1 model transform customer experience for the better

OpenAI o1 models and Hallucinations

While OpenAI has been addressing hallucinations from the start, concerns continue to surround them and their impacts. With the o1 models, the technology company incorporates several advancements aimed at reducing instances where the AI generates incorrect, nonsensical, or fabricated information. These improvements stem from more refined training techniques, better alignment with factual data, and enhanced understanding of context.

Here's how the o1 model reduces hallucinations:

1. Improved Training Data Curation

OpenAI o1 models are trained on a more carefully curated and high-quality dataset than previous models. The dataset includes more accurate, up-to-date information with fewer instances of misleading or incorrect data. By refining the training data, the model is less likely to produce hallucinations since it has a better foundation of reliable knowledge.

OpenAI has reported a reduction in hallucination rates with the o1 models compared to GPT-4. In internal evaluations, hallucinations were reduced by 40-50% in specific knowledge-intensive domains such as scientific, medical, and technical topics.

2. Reinforcement Learning from Human Feedback (RLHF)

The o1 models make extensive use of reinforcement learning from human feedback (RLHF) to refine their responses. In RLHF, human reviewers evaluate the model’s outputs and correct them when necessary. This process helps the model learn from its mistakes, training it to prioritize accurate and contextually relevant information over generating guesses or fabrications. Through this iterative learning process, the model becomes better at avoiding hallucinations.

The o1 models demonstrated a 30% increase in human satisfaction rates during feedback tests. This means that when human reviewers rated the responses, the o1 models were significantly more likely to provide correct, relevant, and coherent information compared to earlier versions.

3. Factual Consistency Mechanisms

OpenAI o1 models integrate mechanisms designed to improve factual consistency. These mechanisms evaluate the generated content against known facts, either during the generation process or through post-processing techniques. By constantly checking against factual baselines, the models can avoid making up information that does not align with reality.

4. Contextual Understanding and Retention

The o1 models have an enhanced ability to understand and retain context throughout longer conversations or content-generation tasks. This reduces hallucinations that often arise from losing track of prior context, which can cause the model to respond with irrelevant or incorrect information. By maintaining a stronger grasp of context, the model generates responses that align more closely with the ongoing discussion or query.

question mark, pile, questions
OpenAI AddressHallucination Concerns with Model o1 (Photo by qimono on Pixabay)

5. Model Calibration

OpenAI has focused on improving the internal "confidence" levels of the o1 models. These models are better at gauging when they are unsure about a response and can signal uncertainty to the user, either by providing more cautious answers or suggesting further verification. This calibration helps avoid overconfident but incorrect outputs, a common cause of hallucinations.

6. Enhanced Multi-step Reasoning

The o1 models incorporate improved multi-step reasoning capabilities. Rather than generating answers based on shallow patterns, the model processes more complex reasoning steps when answering questions. This deeper reasoning helps reduce errors, as the model takes into account various aspects of a problem before providing an answer. In tests involving multi-step reasoning, o1 models outperformed previous iterations by a margin of 15-20% in terms of providing logically consistent and accurate results. This enhancement directly contributes to reducing hallucinations during complex tasks that require deeper reasoning.

7. Knowledge Integration with External Sources

Some iterations of the o1 model can integrate external knowledge bases in real time, allowing the model to pull information from verified sources rather than relying solely on its training data. This dynamic knowledge retrieval can minimize hallucinations, especially when the model is asked about current events or niche topics, by supplementing its responses with accurate, up-to-date information.

8. Self-correction Mechanisms

OpenAI has introduced self-correction mechanisms within the o1 models. These allow the AI to recognize when its initial response may be incomplete or inaccurate and attempt to self-correct before delivering the final answer. This feature helps reduce the likelihood of hallucination by giving the model a second chance to verify its output.

While continual improvements are made to the models, Individuals and organizations should always double-check results and content created before publication.

OpenAI o1 Model and Customer Experience

https://transformidy.com/insight/artificial-intelligence-guide/

Executing Strategy

The integration of these new ChatGPT features can help companies in several key areas of their customer experience (CX) strategy, from improving personalization to driving automation in ways that boost overall satisfaction and loyalty.

  1. Personalized Conversations: The improved NLP in ChatGPT allows for real-time, meaningful dialogue tailored to individual customers. With better memory retention and contextual understanding, companies can deliver more personalized customer experiences. For instance, a customer returning to the same chatbot for support won’t have to explain their issue all over again—ChatGPT can pick up where the conversation left off, ensuring a smoother interaction.
  2. Proactive Engagement: ChatGPT can now engage customers proactively, using data and customer history to initiate conversations. A travel company, for example, can use ChatGPT to reach out to customers with personalized offers based on their past bookings, delivering relevant content and promotions at the perfect moment. This anticipatory approach to customer service helps keep engagement high and customer satisfaction intact.
  3. 24/7 Support with Human-Like Interactions: One of the primary advantages of the latest iteration of ChatGPT is its capability to handle sophisticated queries, freeing human agents to manage more complex issues. With ChatGPT, businesses can provide 24/7 customer support that feels human, resolving concerns at any time of day while maintaining brand consistency.
  4. Enhanced Self-Service Options: ChatGPT’s multi-modal capabilities mean that customers can now engage through text, voice, or even image-based interfaces. For example, in an eCommerce setting, customers can upload pictures of damaged products, and the AI can initiate a support process or offer suggestions for returns or replacements. This reduces friction in the customer journey, leading to more satisfied customers.
  5. Better Feedback Loops: With the chatbot's enhanced ability to retain information and generate reports, businesses can use ChatGPT to gather and analyze customer feedback more effectively. Automatically generating insights based on customer interactions allows companies to adjust their strategies in real time, improving overall experience delivery.

Boosting Customer Engagement

Engagement is at the heart of customer experience, and ChatGPT’s ability to deliver smarter, more personalized interactions gives companies a significant advantage. Customers expect quick, intelligent responses, and the AI’s advanced capabilities allow businesses to meet these expectations more consistently.

  1. Multichannel Interactions: As businesses expand their engagement channels, ChatGPT’s ability to work across platforms (websites, social media, and mobile apps) ensures that customers can interact with a brand in the medium they prefer. With chatbots on multiple platforms, companies can create a unified, omnichannel experience.
  2. Loyalty Program Integration: ChatGPT can also be integrated with loyalty programs, allowing customers to check their points, redeem rewards, or even receive personalized offers in real time. This creates a more interactive and engaging loyalty experience, encouraging repeat purchases and increasing brand affinity.
  3. Gamification and Interactive Engagement: AI-driven customer engagement can also extend into gamification, using quizzes, polls, or interactive content powered by ChatGPT to keep customers engaged. By offering such interactive experiences, brands can encourage more time spent on platforms and deeper engagement with their products and services.
  4. User-Generated Content and Feedback: ChatGPT can facilitate the collection of user-generated content and feedback by prompting customers for reviews, suggestions, or survey responses, making it easier for companies to stay in tune with customer needs and preferences.
https://transformidy.com/insight/maximizing-the-chatbots/

Driving Revenue Generation

One of the most significant impacts of ChatGPT’s latest developments is on revenue generation. By streamlining processes, increasing personalization, and improving the customer journey, businesses can expect to see a direct boost in sales.

  1. Targeted Recommendations: With enhanced data processing capabilities, ChatGPT can offer personalized product recommendations to customers based on their purchase history or browsing behavior. This creates more opportunities for upselling and cross-selling, driving higher average order values.
  2. Streamlined Sales Processes: AI chatbots can now assist customers through each stage of the sales funnel, from initial inquiry to final purchase. ChatGPT can answer questions about products, handle payment issues, and even recommend complementary products, reducing the chances of cart abandonment.
  3. Automated Order Management: Companies can also use ChatGPT to automate order management tasks such as tracking shipments, processing returns, or handling cancellations. This removes bottlenecks in the sales process, allowing businesses to focus on growth and customer acquisition rather than manual backend processes.
  4. Dynamic Pricing and Offers: By analyzing customer behavior and market trends in real time, ChatGPT can suggest dynamic pricing strategies or offer tailored discounts to specific segments. This real-time adaptability can lead to increased conversions and customer satisfaction.

Strengthening Partnerships

Companies that leverage AI like ChatGPT can improve not just customer-facing processes but also partnerships and B2B relationships. The enhanced data processing and reporting capabilities allow for more transparency and collaboration between business partners.

  1. Shared Customer Insights: ChatGPT can generate detailed reports based on customer interactions, which can be shared with partners to better understand joint audiences. For example, a brand partnership in retail might use AI-generated insights to co-create marketing campaigns tailored to a shared target demographic.
  2. Automated Partner Communication: By integrating ChatGPT into communication platforms, businesses can streamline communication with partners, automating routine updates or status reports while maintaining a human-like interaction quality.
  3. Improved Coordination for Joint Initiatives: ChatGPT’s ability to analyze data and offer real-time recommendations can also help companies and partners better coordinate their joint initiatives, ensuring alignment on campaigns, offers, and strategies that resonate with mutual customer bases.
  4. Enhanced Collaboration through AI Tools: ChatGPT’s multi-modal capabilities can support collaboration between partners, facilitating not only textual but also image-based and video-based communication, allowing companies to work more effectively on creative initiatives.

Building Trust In the Process

There is an inherent trust issue with using AI models in building products and services. To improve the trust equation, organizations need to prioritize transparency about AI usage, ensuring customers understand how their data is collected, processed, and used. Clear communication about the AI's capabilities and limitations helps prevent misunderstandings and builds trust.

https://transformidy.com/insight/trust-balancing-experiences-costs-growth/

Additionally, organizations should invest in robust data privacy and security measures to protect customer information and prevent misuse, test and monitor how AI models are used, and manage any hallucinations. By fostering trust from the beginning, organizations can build a system to manage the use of AI and use the technology to implement better customer experiences, strengthen brand loyalty, and enhance overall business success.

Transform For The Better

The release of OpenAI o1 models marks a new era in customer experience, where AI-driven personalization, engagement, and efficiency become the standard. By adopting these models, companies can revolutionize how they execute their customer experience strategies, build deeper connections with customers, drive revenue, and strengthen partnerships. The opportunities are limitless, and businesses that embrace this new wave of AI technology are poised to lead the way into the future.

This transformation is not just about improving processes—it's about creating a business culture that puts the customer first, enhances employee capabilities, and drives meaningful, long-lasting success. The integration of OpenAI o1 models represents an essential step in transforming for the better. Companies that seize this opportunity will not only thrive but set new benchmarks for what exceptional customer experience looks like in the modern age.

How Can We Help?

Transformidy is available to assist in helping you understand OpenAI, ChatGPT, and the o1 model as part of your company’s customer experience strategy. We are also available to assess how effectiveness of your company’s experience strategy in generating engagement, satisfaction, and business growth.

Contact us or set up a 30-minute complimentary consultation for more information on our services, insights, or showcases. We look forward to hearing from you.

FAQ

How many consumers expect an explanation for an AI-made decision?

According to Zendesk's 2026 CX Trends research, 95% of consumers now expect to understand the 'why' behind an automated decision. Demand for AI transparency specifically rose 63% year over year, indicating this expectation is accelerating, not holding steady.

Why did this expectation rise so quickly?

Reasoning-capable AI models becoming mainstream in customer-facing tools normalized the idea that an AI system can show its work, not just produce an answer. Once customers experienced tools that explained their reasoning, an unexplained automated decision, a denied claim, a rejected order, a flagged transaction, started to feel like a withheld answer rather than a normal limitation of the technology.

Does accuracy alone satisfy this expectation?

No. Zendesk's research frames the finding specifically around clarity, not correctness: the quoted conclusion is that 'it's clarity that drives trust and loyalty in CX engagements,' distinct from whether the underlying decision was accurate. A correct decision delivered without explanation can still damage trust if the customer cannot see why it was made.

What should a business do if its AI systems cannot currently explain their reasoning?

Treat explainability as a customer experience requirement, not just a technical or compliance one, and prioritize it in AI vendor selection and internal system design. A 63% year-over-year rise in transparency demand suggests this gap will keep widening for any organization that does not close it deliberately.